Summarization on EMAILSUM long 1.0 (test)
45.98ROUGE-1 (R1)Oracle
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Oracle2021.07 | 45.98 | 15.49 | 32.4 | 42.14 | 26.31 | |
| SemiSupw3cBase Model=T5-base, Learning Paradigm=Semi-supervised, Unlabeled Data Source=W3C2021.07 | 44.56 | 14.6 | 31.38 | 40.73 | 32.81 | |
| Hier. T5baseBase Model=T5-base, Model Architecture=Hierarchical2021.07 | 44.5 | 14.53 | 30.89 | 40.22 | 32.3 | |
| CNNDMpreBase Model=T5-base, Transfer Learning Source Dataset=CNN/DM, Training Strategy=Pre-finetuning2021.07 | 44.15 | 14.2 | 30.84 | 40.21 | 32.53 | |
| SemiSuptogetherBase Model=T5-base, Learning Paradigm=Semi-supervised, Unlabeled Data Source=Together (W3C + Avocado)2021.07 | 44.08 | 14.06 | 31.17 | 40.67 | 32.3 | |
| SemiSupavocadoBase Model=T5-base, Learning Paradigm=Semi-supervised, Unlabeled Data Source=Avocado2021.07 | 43.83 | 14.61 | 31.21 | 40.52 | 32.71 | |
| T5baseBase Model=T5-base, Learning Paradigm=Supervised, Model Architecture=Non-hierarchical2021.07 | 43.81 | 14.08 | 30.47 | 39.88 | 32.09 | |
| CRD3preBase Model=T5-base, Transfer Learning Source Dataset=CRD3, Training Strategy=Pre-finetuning2021.07 | 43.6 | 13.93 | 30.49 | 39.97 | 31.53 | |
| XSumpreBase Model=T5-base, Transfer Learning Source Dataset=XSum, Training Strategy=Pre-finetuning2021.07 | 43.48 | 13.82 | 30.14 | 39.8 | 31.6 | |
| CNNDMjointBase Model=T5-base, Transfer Learning Source Dataset=CNN/DM, Training Strategy=Joint-training2021.07 | 43.28 | 12.37 | 28.84 | 39.39 | 29.95 | |
| SAMSumjointBase Model=T5-base, Transfer Learning Source Dataset=SAMSum, Training Strategy=Joint-training2021.07 | 42.96 | 13.44 | 29.99 | 39.54 | 31.82 | |
| SAMSumpreBase Model=T5-base, Transfer Learning Source Dataset=SAMSum, Training Strategy=Pre-finetuning2021.07 | 42.83 | 13.54 | 30 | 39.13 | 31.82 | |
| CRD3jointBase Model=T5-base, Transfer Learning Source Dataset=CRD3, Training Strategy=Joint-training2021.07 | 42.81 | 12.96 | 29.35 | 39.32 | 32.14 | |
| XSumjointBase Model=T5-base, Transfer Learning Source Dataset=XSum, Training Strategy=Joint-training2021.07 | 42.36 | 11.85 | 28.23 | 38.31 | 29.22 | |
| Fast Abs RL2021.07 | 39.35 | 10.58 | 27.01 | 36.51 | 10.03 | |
| Lead-1-Email2021.07 | 35.71 | 8.69 | 24.7 | 32.13 | 16.93 | |
| BertSumExt2021.07 | 30.23 | 7.08 | 19.59 | 26.68 | 7.78 | |
| TextRank2021.07 | 28.42 | 6.2 | 19.08 | 25.19 | 5.67 | |
| Lead-12021.07 | 19.75 | 4.84 | 14.24 | 16.88 | 6.87 |